Which SRS is computational more effective to store continental scale data? (given some conditions)

Question

For a Precision Agriculture application I need to make calculations in UTM projection. The data will be stored in Postgis and is made of soil samples results (point data) and field boundaries of many farms. But the farms are scattered through many utm zones (ex. a farm is in Z23 and the other is in Z22, so I can't store the data in UTM, it needs to be stored in a SRS without zones.

Since the calculations will be made in UTM the data will always need to be converted back and forth to the storage SRS.

So which of those inputs is cheaper in terms of computations of SRS conversion when the output SRS is UTM Wgs84:
-Geometry vs. Geography Postgis types
-Geometry projected (ex World Mercator) vs. Unprojected (latlon)

I think that it involves the internal Proj4 math but I couldn't understand it myself...

Answer 1

I'm not an expert on the subject, so feedback from others is welcome, but I think you should store in global geographic CRS - i.e. lat / lon in decimal degrees. Here's why (and this is the bit that could be wrong / inaccurate):

When converting between projected CRSs you have to first convert to geographic and then to the target CRS. Each CRS must, as part of its definition, define how to convert coordinates to and from geographic, but doesn't need to define conversions to any other CRS. Therefore storing in a geographic CRS means only 1 operation, not 2.